#!/usr/bin/env python3 import os import re from pathlib import Path from typing import List BASE_URL = "https://huggingface.co/csukuangfj/sherpa-onnx-apk/resolve/main/" from dataclasses import dataclass @dataclass class APK: major: int minor: int patch: int arch: str name1: str name2: str def __init__(self, s): # sherpa-onnx-1.10.27-arm64-v8a-speaker-diarization-pyannote_audio-3dspeaker.apk s = str(s).split("/")[-1] split = s.split("-") self.major, self.minor, self.patch = list(map(int, split[2].split("."))) self.arch = split[3] self.name1 = split[-2] self.name2 = split[-1] if "arm" in s: self.arch += "-" + split[4] if "armeabi" in self.arch: self.arch = "y" + self.arch if "arm64" in self.arch: self.arch = "z" + self.arch def sort_by_apk(x): x = APK(x) return (x.major, x.minor, x.patch, x.arch, x.name1, x.name2) def get_all_files(d_list: List[str], suffix: str) -> List[str]: if isinstance(d_list, str): d_list = [d_list] min_major = 1 min_minor = 9 min_patch = 10 ss = [] for d in d_list: for root, _, files in os.walk(d): for f in files: if f.endswith(suffix): major, minor, patch = list(map(int, f.split("-")[2].split("."))) if major >= min_major and minor >= min_minor and patch >= min_patch: ss.append(os.path.join(root, f)) ans = sorted(ss, key=sort_by_apk, reverse=True) return list(map(lambda x: BASE_URL + str(x), ans)) def to_file(filename: str, files: List[str]): content = r"""

APKs for Speaker diarization

This page lists the speaker diarization APKs for sherpa-onnx, one of the deployment frameworks of the Next-gen Kaldi project.
The name of an APK has the following rule: where
Note: Please see https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-segmentation-models for a list of supported speaker segmentation models.
Note: Please see https://github.com/k2-fsa/sherpa-onnx/releases/tag/speaker-recongition-models for a list of supported speaker embedding extraction models.

Note about the license The code of Next-gen Kaldi is using Apache-2.0 license. However, we support models from different frameworks. Please check the license of your selected model.

Note about the build script You can find the script for building the APKs at https://github.com/k2-fsa/sherpa-onnx/blob/master/scripts/apk/build-apk-speaker-diarization.sh

APK Speaker segmentation model Speaker embedding extraction model
sherpa-onnx-x.y.z-arm64-v8a-speaker-diarization-pyannote_audio-3dspeaker.apk sherpa-onnx-pyannote-segmentation-3-0.tar.bz2

It is converted from https://huggingface.co/pyannote/segmentation-3.0		 3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx

It is converted from https://github.com/alibaba-damo-academy/3D-Speaker
sherpa-onnx-x.y.z-arm64-v8a-speaker-diarization-pyannote_audio-nemo.apk sherpa-onnx-pyannote-segmentation-3-0.tar.bz2

It is converted from https://huggingface.co/pyannote/segmentation-3.0		 nemo_en_titanet_small.onnx

It is converted from https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/titanet_small
sherpa-onnx-x.y.z-arm64-v8a-speaker-diarization-revai_v1-3dspeaker.apk sherpa-onnx-reverb-diarization-v1.tar.bz2

It is converted from https://huggingface.co/Revai/reverb-diarization-v1		 3dspeaker_speech_eres2net_base_sv_zh-cn_3dspeaker_16k.onnx

It is converted from https://github.com/alibaba-damo-academy/3D-Speaker
sherpa-onnx-x.y.z-arm64-v8a-speaker-diarization-revai_v1-nemo.apk sherpa-onnx-reverb-diarization-v1.tar.bz2

It is converted from https://huggingface.co/Revai/reverb-diarization-v1		 nemo_en_titanet_small.onnx

It is converted from https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/titanet_small


""" if "-cn" not in filename: content += """ For Chinese users, please visit this address, which replaces huggingface.co with hf-mirror.com

中国用户, 请访问这个地址

""" with open(filename, "w") as f: print(content, file=f) for x in files: name = x.rsplit("/", maxsplit=1)[-1] print(f'{name}
', file=f) def main(): apk = get_all_files("speaker-diarization", suffix=".apk") to_file("./apk-speaker-diarization.html", apk) # for Chinese users apk2 = [] for a in apk: a = a.replace("huggingface.co", "hf-mirror.com") a = a.replace("resolve", "blob") apk2.append(a) to_file("./apk-speaker-diarization-cn.html", apk2) if __name__ == "__main__": main()